特征矢量优化的滚动轴承故障诊断

Application of the Optimized Feature Vectors for Fault Diagnosis of Rolling Element Bearings

为提取小波包频带中的有效故障信息,基于Fisher线性测度提出一种新的特征矢量优化方法。轴承振动信号经小波包分解后,各子频带数据片段的能量值作为参数构建特征矢量。使用差异性和相似性优化相结合方法,分别选出不同轴承状态下Fisher距离较大的小波包频带,以及同种轴承状态下Fisher距离最小的频带,提取出易于区分不同轴承状态的故障信息。故障辨识使用连续型隐马尔可夫模型,在3种故障程度下实现了轴承正常状态、滚动体故障、内圈和外圈故障的有效判别,辨识精度大于94%。比较实验表明文中方法的辨识精度优于文献方法。

A new approach of feature vector optimization for extracting the effective fault information was presented using Fisher linear distance. Firstly,the vibration signals were decomposed into the sub-bands with the wavelet packet transform,and the energy of which was used to construct the feature vectors. Then,the methods of difference and similarity optimization were applied to select the sub-bands which have greater Fisher distance between the different bearing statuses,and meantime has the minimal Fisher distance within the same bearing status. The fault identification applied the continuous hidden Markov models,which successfully identified normal status,ball fault,inner race fault and outer race fault in three kinds of fault severities,and the identification accuracy was greater than 94%. The result of compared experiments showed the identification accuracy of the presented method was better than the reference method.